Background: Now that it is possible to efficiently classify and save tissue images of laboratory animals using wholeslide imaging, many diagnostic models are being developed through transfer learning with Convolutional Neural Network (CNN). In this study, transfer learning was performed to gain toxicopathological knowledge using CNN models such as InceptionV3 and Xception. For the classification of tubular basophilia and mineralization, two representative background lesions that commonly occur in toxicological studies, accuracies of diagnosis were compared using MobileNetV2, Xception and InceptionV3. For the simultaneous detection of the two lesions, the accuracy was analysed using You Only Look Once version 4 (YOLOv4). Results: The accuracy of the classification models was as follows: MobileNetV2 (epoch 50, accuracy: 98.57%) > Xception (epoch 70, accuracy: 97.47%) > InceptionV3 (epoch 70, accuracy: 89.62%). In the case of object detection, the accuracy of YOLOv4 was 98.62% at epoch 3000. Conclusions Among the classification models, MobileNetV2 had the best accuracy despite applying a lower epoch than InceptionV3 and Xception. The object detection model, YOLOv4, accurately and simultaneously diagnosed tubular basophilia and mineralization, with an accuracy of 98.62% at epoch 3000.

Background/Aims: Useful biomarkers for metabolic syndrome have been insufficient. We investigated the performance of serum milk fat globule-EGF factor-8 (MFG-E8), the key mediator of inflammatory pathway, in diagnosis of metabolic syndrome. Methods: Subjects aged between 30 and 64 years were prospectively enrolled in the Seoul Metabolic Syndrome cohort. Serum MFG-E8 levels were measured at baseline. Results: A total of 556 subjects were included, comprising 279 women (50.2%) and 277 men (49.8%). Metabolic syndrome was diagnosed in 236 subjects (42.4%), and the mean MFG-E8 level of subjects with metabolic syndrome was significantly higher than that of subjects without metabolic syndrome (P<0.001). MFG-E8 level was significantly correlated with all metabolic syndrome components and pulse wave velocity (all P<0.05). Subjects were categorized into two groups according to the best MFG-E8 cut-off value as follows: group 1, MFG-E8 level <4,745.1 pg/mL (n=401, 72.1%); and group 2, MFG-E8 level ≥4,745.1 (n=155, 27.9%). At baseline, metabolic syndrome in group 2 was significantly more prevalent than in group 1 (63.9% vs. 34.2%, P<0.001). During median follow-up of 17 months, metabolic syndrome developed in 122 (38.1%) subjects among 320 subjects without it at baseline. The incidence of metabolic syndrome in group 2 was significantly higher than that in group 1 (55.4% vs. 34.5%, P=0.003). On multivariate analysis, MFG-E8 level ≥4,745.1 pg/mL was an independent predictor for diagnosis and development of metabolic syndrome after adjusting other factors (all P<0.05). Conclusions Serum MFG-E8 level is a potent biomarker for the screening and prediction of metabolic syndrome.

Background/Aims: Most prognostic prediction models for patients with liver cirrhosis include serum total bilirubin (TB) level as a component. This study investigated prognostic performance of serum direct bilirubin (DB) and developed new DB level-based prediction models for cirrhosis. Methods: A total of 983 hospitalized patients with liver cirrhosis were included. DB-Model for End-Stage Liver Disease (MELD) score was calculated using MELD score formula, with serum DB level replacing TB level. Results: Mean age of study population was 56.1 years. Alcoholic liver disease was the most frequent underlying condition (471 patients, 47.9%). Within 6 months, 144 patients (14.6%) died or received liver transplantation due to severe liver dysfunction. The area under the receiver operating characteristic curve (AUROC) for prediction of 6-month mortality with DB level was significantly higher than that with TB level (p<0.001). The AUROC of DB-MELD score for prediction of 6-month mortality was significantly higher than that of MELD score (p<0.001). Patients were randomly divided into training (n=492) and validation (n=491) cohorts. A new prognostic prediction model, “Direct Bilirubin, INR, and Creatinine” (DiBIC) score, was developed based on the most significant predictors of 6-month mortality. In training set, AUROC of DiBIC score for prediction of 6-month mortality was 0.892, which was significantly higher than that of the MELD score (0.875, p=0.017), but not different from that of DB-MELD score (0.886, p=0.272). Similar results were observed in validation set. Conclusions New prognostic models, DB-MELD and DiBIC scores, have good prognostic performance in liver cirrhosis patients, outperforming other currently available models.

Background/Aims: We investigated changes in recurrence rates and significant recurrence predictors over time after complete cure of hepatocellular carcinoma (HCC). Methods: A total of 1,491 patients with first-time diagnosis of Barcelona Clinic Liver Cancer stage A HCC, completely cured by treatment between 2007 and 2016, were recruited from two Korean tertiary institutes. Results: The mean age of the population (1,144 men and 347 women) was 58.6 years. Of the total population, 914 patients (61.3%) had liver cirrhosis. Nine-hundred and forty-one (63.1%) and 550 (36.9%) patients were treated with surgical resection and radiofrequency ablation (RFA), respectively. One-year cumulative incidences of HCC recurrence were 14.3%, 9.9%, and 5.1% from the time of treatment, 3 years after treatment, and 5 years after treatment, respectively. Upon multivariate analysis, multiple tumors, maximal tumor size ≥3 cm, and high Model for End-Stage Liver Disease scores were independently associated with increased HCC recurrence risk from the time of treatment and 1 and 2 years after curative treatment (all p<0.05, except for maxi-mal tumor size ≥3 cm for recurrence 2 years after treatment). Meanwhile, liver cirrhosis and RFA were independently associated with the increased HCC recurrence risk for almost all time points (liver cirrhosis: all p<0.05; RFA: all p<0.005 except for recurrence from 5 years after treatment). Conclusions The recurrence rate of HCC after curative treatment gradually decreased over time. Two years after treatment, when tumor-related factors lose their prognostic implications, may be used as a cutoff to define the boundary between early and late recurrence of HCC.

Background/Aims: Most prognostic prediction models for patients with liver cirrhosis include serum total bilirubin (TB) level as a component. This study investigated prognostic performance of serum direct bilirubin (DB) and developed new DB level-based prediction models for cirrhosis. Methods: A total of 983 hospitalized patients with liver cirrhosis were included. DB-Model for End-Stage Liver Disease (MELD) score was calculated using MELD score formula, with serum DB level replacing TB level. Results: Mean age of study population was 56.1 years. Alcoholic liver disease was the most frequent underlying condition (471 patients, 47.9%). Within 6 months, 144 patients (14.6%) died or received liver transplantation due to severe liver dysfunction. The area under the receiver operating characteristic curve (AUROC) for prediction of 6-month mortality with DB level was significantly higher than that with TB level (p<0.001). The AUROC of DB-MELD score for prediction of 6-month mortality was significantly higher than that of MELD score (p<0.001). Patients were randomly divided into training (n=492) and validation (n=491) cohorts. A new prognostic prediction model, “Direct Bilirubin, INR, and Creatinine” (DiBIC) score, was developed based on the most significant predictors of 6-month mortality. In training set, AUROC of DiBIC score for prediction of 6-month mortality was 0.892, which was significantly higher than that of the MELD score (0.875, p=0.017), but not different from that of DB-MELD score (0.886, p=0.272). Similar results were observed in validation set. Conclusions New prognostic models, DB-MELD and DiBIC scores, have good prognostic performance in liver cirrhosis patients, outperforming other currently available models.

Background/Aims: We investigated changes in recurrence rates and significant recurrence predictors over time after complete cure of hepatocellular carcinoma (HCC). Methods: A total of 1,491 patients with first-time diagnosis of Barcelona Clinic Liver Cancer stage A HCC, completely cured by treatment between 2007 and 2016, were recruited from two Korean tertiary institutes. Results: The mean age of the population (1,144 men and 347 women) was 58.6 years. Of the total population, 914 patients (61.3%) had liver cirrhosis. Nine-hundred and forty-one (63.1%) and 550 (36.9%) patients were treated with surgical resection and radiofrequency ablation (RFA), respectively. One-year cumulative incidences of HCC recurrence were 14.3%, 9.9%, and 5.1% from the time of treatment, 3 years after treatment, and 5 years after treatment, respectively. Upon multivariate analysis, multiple tumors, maximal tumor size ≥3 cm, and high Model for End-Stage Liver Disease scores were independently associated with increased HCC recurrence risk from the time of treatment and 1 and 2 years after curative treatment (all p<0.05, except for maxi-mal tumor size ≥3 cm for recurrence 2 years after treatment). Meanwhile, liver cirrhosis and RFA were independently associated with the increased HCC recurrence risk for almost all time points (liver cirrhosis: all p<0.05; RFA: all p<0.005 except for recurrence from 5 years after treatment). Conclusions The recurrence rate of HCC after curative treatment gradually decreased over time. Two years after treatment, when tumor-related factors lose their prognostic implications, may be used as a cutoff to define the boundary between early and late recurrence of HCC.

Background/Aims: To prevent the perinatal transmission of hepatitis B virus (HBV) from mother to child, administration of an antiviral agent during pregnancy has been attempted in women who are either hepatitis B e antigen positive or have a high viral load. In this systematic review and meta-analysis with randomized controlled trials, we analyzed the efficacy and safety of tenofovir disoproxil fumarate (TDF) in preventing the perinatal transmission of HBV in pregnant women who have high HBV DNA titers. Methods: Multiple comprehensive databases (PubMed, EMBASE, and Cochrane databases) were searched for studies evaluating the efficacy of TDF for the prevention of perinatal transmission of HBV. Results: Two studies (one open label study and one double blind study) were included and analyzed. Intention-to-treat analysis (527 pregnancies) showed that the preventive effect of TDF was not significant (odds ratio [OR], 0.53; 95% confidence interval[CI], 0.13 to 2.17; p = 0.38, I2 = 81%). However, the per-protocol analysis showed that TDF significantly reduced perinatal transmission (OR, 0.10; 95% CI, 0.01 to 0.77; p = 0.03, I2 = 0%). There was no significant difference between the TDF group and the control group with respect to maternal and fetal safety outcomes. Conclusions In pregnant women who have high HBV DNA titers, TDF can reduce the perinatal transmission from mother to child without significant adverse events.

Background/Aims: Useful biomarkers for metabolic syndrome have been insufficient. We investigated the performance of serum milk fat globule-EGF factor-8 (MFG-E8), the key mediator of inflammatory pathway, in diagnosis of metabolic syndrome. Methods: Subjects aged between 30 and 64 years were prospectively enrolled in the Seoul Metabolic Syndrome cohort. Serum MFG-E8 levels were measured at baseline. Results: A total of 556 subjects were included, comprising 279 women (50.2%) and 277 men (49.8%). Metabolic syndrome was diagnosed in 236 subjects (42.4%), and the mean MFG-E8 level of subjects with metabolic syndrome was significantly higher than that of subjects without metabolic syndrome (P<0.001). MFG-E8 level was significantly correlated with all metabolic syndrome components and pulse wave velocity (all P<0.05). Subjects were categorized into two groups according to the best MFG-E8 cut-off value as follows: group 1, MFG-E8 level <4,745.1 pg/mL (n=401, 72.1%); and group 2, MFG-E8 level ≥4,745.1 (n=155, 27.9%). At baseline, metabolic syndrome in group 2 was significantly more prevalent than in group 1 (63.9% vs. 34.2%, P<0.001). During median follow-up of 17 months, metabolic syndrome developed in 122 (38.1%) subjects among 320 subjects without it at baseline. The incidence of metabolic syndrome in group 2 was significantly higher than that in group 1 (55.4% vs. 34.5%, P=0.003). On multivariate analysis, MFG-E8 level ≥4,745.1 pg/mL was an independent predictor for diagnosis and development of metabolic syndrome after adjusting other factors (all P<0.05). Conclusions Serum MFG-E8 level is a potent biomarker for the screening and prediction of metabolic syndrome.